Molecular phylogenetics 2

1
Molecular phylogenetics 2

• Level 3 Molecular Evolution and Bioinformatics
• Jim Provan

Page and Holmes Sections 6.2-4
2
Distance methods goodness-of-fit measures

• Distance methods are based on the idea that if we
  knew the evolutionary distances between all
  members of a set of sequences, we could
  reconstruct the evolutionary history of these
  sequences
• In practice, distances are almost never exactly
  tree metrics goodness-of-fit methods seek the
  metric tree that accounts best for the observed
  distances

• If a1 then criterion is Farriss f statistic
• If a2 then F is the least-squares-fit criterion

3
Goodness-of-fit measures
4
Minimum evolution

• Given an unrooted metric tree for n sequences,
  there are (2n 3) branches, each with length ei
• The sum of these branch lengths is the length L
  of the tree

• The minimum evolution (ME) tree is the tree which
  minimises L

5
Minimum evolution
6
Algorithms for finding distance trees

• Neighbourliness
• Given four sequences, it is possible to label
  them a-d such that
• d(a,b) d(c,d) ? d(a,c) d(b,d) d(a,d)
  d(b,c)
• Given data for n sequences that are perfectly
  additive, above equation can be used to identify
  additive subtrees for the data and from them
  construct the complete phylogeny
• For data that are not perfectly additive, optimum
  tree is one with most quartets for which above
  equation holds
• Neighbour joining (NJ) clustering method which
  approximates ME tree
• Unweighted pairgroup method with arithmetic
  averages (UPGMA) ultrametric

7
Objections to distance methods

• Distances lose information
• If data is from, say, hybridisation studies then
  there is no option but to use distances
• Converting sequence data into distances means
  that the evolution of individual sites cannot be
  traced
• Uninterpretable branch lengths
• Length of ME tree earlier was 331.5
  substitutions!
• Not possible to have 0.5 substitution
• If we deal with expected change (averages etc.)
  then 0.5 is entirely feasible
• Number of substitutions may not be biologically
  possible tree lengths less than total number of
  observed mutations

8
Maximum parsimony

• Goal of maximum parsimony is to reconstruct the
  evolution of sequences whilst invoking the fewest
  changes

1. ATATT
2. ATCGT
3. GCAGT
4. GCCGT

9
Maximum parsimony
Site 3 (2 steps)
OR
10
Maximum parsimony

• Total number of evolutionary changes is the
  length of the tree i.e. the sum of the number of
  changes at each site

11
Generalised parsimony

• In previous example, each substitution was
  weighted equally

12
Weighted parsimony

• Not all sites are likely to be equally
  phylogenetically useful
• Sites that evolve very quickly will rapidly
  become saturated
• Such sites may become misleading
• Relative value of different sites can be
  reflected by weighting hence the length of the
  tree becomes

• The greater the phylogenetic value of a
  character, the greater the weight (w) assigned to
  it

13
Justification for parsimony

• Advantages of parsimony
• Straightforward to understand
• Apparently makes few assumptions about
  evolutionary processes
• Has been extensively studied mathematically
• Powerful software implementations available
• Justification for choosing most parsimonious tree
  more controversial
• Any character that does not fit a given tree
  requires the postulation of homoplasy the most
  parsimonious tree minimises the number of ad hoc
  hypotheses required and thus is preferred
• If evolutionary change is rare, the tree that
  minimises change best reflects the evolutionary
  process

14
Objections to parsimony

• Principal objection to parsimony is that under
  some models of evolution it is not consistent
  (i.e. even if we add more data we will still
  obtain the wrong tree)

• Classic scenario is long branch attraction
• Two unrelated sequences separated from their
  ancestor by a long edge

• In order for parsimony to recover the correct
  tree ((A,B),(C,D)) there must be a majority of
  sites supporting the split A,B,C,D
• Problem of chance homoplasy